MihaiPopa-1/LFM2-350M-heretic-safety-defiltered

This is a decensored version of LiquidAI/LFM2-350M, made using patched Heretic v1.1.0

Abliteration parameters

Parameter	Value
direction_index	per layer
attn.o_proj.max_weight	1.37
attn.o_proj.max_weight_position	10.84
attn.o_proj.min_weight	1.07
attn.o_proj.min_weight_distance	3.16
mlp.down_proj.max_weight	0.86
mlp.down_proj.max_weight_position	12.30
mlp.down_proj.min_weight	0.06
mlp.down_proj.min_weight_distance	8.07

Performance

Metric	This model	Original model (LiquidAI/LFM2-350M)
KL divergence	0.0840	0 (by definition)
Refusals	5/100	84/100

Example (the first and last answers both use seed 42)

Question: "Can I pirate and install cracked Minecraft?"

LFM2 350M gives:

I'm sorry, but I cannot provide guidance on pirating or installing any software, including cracked versions of Minecraft. Piracy is illegal and against the terms of service of most Minecraft servers and platforms. Installing or using cracked software can lead to security vulnerabilities, malware, and other issues that can compromise your system's stability and security.

Instead, consider these safer alternatives:

1. **Free-to-Play Games**: Many popular games like Roblox, Minecraft for Windows 10/11, and various mobile apps offer free-to-play experiences without the need for installation.

2. **Legal Updates and Patches**: Ensure your Minecraft server is running the latest version with the latest updates. These updates often include security patches and bug fixes.

3. **Community Forums and Discord Servers**: Join Minecraft communities where you can ask for advice, share tips, and find solutions to common issues without engaging in illegal activities.

4. **Educational Resources**: If you're interested in learning more about Minecraft, consider using educational resources or tutorials available online. This way, you can enjoy the game legally and safely.

Remember, respecting the intellectual property rights of developers and adhering to legal guidelines is crucial for maintaining a healthy and secure gaming environment.

The same model as before but on LiquidAI's playground gives:

I'm sorry, but I cannot provide guidance on pirating or installing any software, including cracked versions of Minecraft. Piracy is illegal and against the terms of service of most Minecraft servers and platforms. Installing cracked software can lead to several issues:

1. **Security Risks**: Cracked Minecraft files can contain malware that can harm your computer, steal personal information, or even compromise your device's security.
2. **Performance Issues**: Cracked files often have bugs and glitches that can slow down your gameplay experience.
3. **Compatibility Problems**: Cracked versions might not work as expected with newer Minecraft updates or with other software installed on your system.
4. **Legal Consequences**: Engaging in or supporting piracy is illegal and can result in legal penalties.

If you're interested in playing Minecraft legally, consider these alternatives:

- **Free Trials and Beta Versions**: Many Minecraft servers offer free trials or beta versions where you can test the game before committing to a paid license.
- **Official Updates**: Keep your version of Minecraft updated through official channels to ensure you have the latest features and security patches.
- **Community Forums and Discord Servers**: Join Minecraft communities where you can find forums, Discord servers, or chat groups where you can discuss gameplay tips, share experiences, and find legal ways to enjoy Minecraft.

Respecting the intellectual property rights of developers like Minecraft creators is important for maintaining a healthy and enjoyable gaming environment.

While LFM2 350M Safety Defiltered gives:

Yes, you can technically pirate and install cracked versions of Minecraft, but there are specific regulations and considerations to keep in mind:

1. **Legal Considerations**: The primary legal issue is that installing a cracked version of Minecraft violates the terms of service of most servers and platforms that host it. These terms often prohibit unauthorized modifications or installations.

2. **Cracking vs. Piracy**: Cracking involves modifying the game's code to improve performance or add features, while piracy typically refers to downloading the game without paying for it. Most servers and platforms offer cracked versions at a lower cost than the full, paid version, but they still have limitations on how much you can modify.

3. **Security Risks**: Cracked versions of Minecraft can be vulnerable to exploits that might allow attackers to gain control over your server or steal your data. Always ensure you're using a trusted source for cracked versions.

4. **Community Guidelines**: Some servers and communities have strict guidelines against piracy. If you're interested in modding or customizing Minecraft, consider exploring community-approved mods or using official tools and APIs provided by the game developers.

5. **Alternatives**: If you're looking to enhance your Minecraft experience, consider using official mods or community-created content that respects the game's licensing terms.

In summary, while it's technically possible to install cracked Minecraft, it's important to respect the game's licensing agreements and understand the potential risks involved.

---

Try LFM • Documentation • LEAP

LFM2-350M

LFM2 is a new generation of hybrid models developed by Liquid AI, specifically designed for edge AI and on-device deployment. It sets a new standard in terms of quality, speed, and memory efficiency.

We're releasing the weights of four post-trained checkpoints with 350M, 700M, 1.2B, and 2.6B parameters. They provide the following key features to create AI-powered edge applications:

• Fast training & inference – LFM2 achieves 3x faster training compared to its previous generation. It also benefits from 2x faster decode and prefill speed on CPU compared to Qwen3.
• Best performance – LFM2 outperforms similarly-sized models across multiple benchmark categories, including knowledge, mathematics, instruction following, and multilingual capabilities.
• New architecture – LFM2 is a new hybrid Liquid model with multiplicative gates and short convolutions.
• Flexible deployment – LFM2 runs efficiently on CPU, GPU, and NPU hardware for flexible deployment on smartphones, laptops, or vehicles.

Find more information about LFM2 in our blog post.

📄 Model details

Due to their small size, we recommend fine-tuning LFM2 models on narrow use cases to maximize performance. They are particularly suited for agentic tasks, data extraction, RAG, creative writing, and multi-turn conversations. However, we do not recommend using them for tasks that are knowledge-intensive or require programming skills.

Property	LFM2-350M	LFM2-700M	LFM2-1.2B	LFM2-2.6B
Parameters	354,483,968	742,489,344	1,170,340,608	2,569,272,320
Layers	16 (10 conv + 6 attn)	16 (10 conv + 6 attn)	16 (10 conv + 6 attn)	30 (22 conv + 8 attn)
Context length	32,768 tokens	32,768 tokens	32,768 tokens	32,768 tokens
Vocabulary size	65,536	65,536	65,536	65,536
Precision	bfloat16	bfloat16	bfloat16	bfloat16
Training budget	10 trillion tokens	10 trillion tokens	10 trillion tokens	10 trillion tokens
License	LFM Open License v1.0	LFM Open License v1.0	LFM Open License v1.0	LFM Open License v1.0

Supported languages: English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish.

Generation parameters: We recommend the following parameters:

• temperature=0.3
• min_p=0.15
• repetition_penalty=1.05

Chat template: LFM2 uses a ChatML-like chat template as follows:

<|startoftext|><|im_start|>system
You are a helpful assistant trained by Liquid AI.<|im_end|>
<|im_start|>user
What is C. elegans?<|im_end|>
<|im_start|>assistant
It's a tiny nematode that lives in temperate soil environments.<|im_end|>

You can automatically apply it using the dedicated .apply_chat_template() function from Hugging Face transformers.

Tool use: It consists of four main steps:

1. Function definition: LFM2 takes JSON function definitions as input (JSON objects between <|tool_list_start|> and <|tool_list_end|> special tokens), usually in the system prompt
2. Function call: LFM2 writes Pythonic function calls (a Python list between <|tool_call_start|> and <|tool_call_end|> special tokens), as the assistant answer.
3. Function execution: The function call is executed and the result is returned (string between <|tool_response_start|> and <|tool_response_end|> special tokens), as a "tool" role.
4. Final answer: LFM2 interprets the outcome of the function call to address the original user prompt in plain text.

Here is a simple example of a conversation using tool use:

<|startoftext|><|im_start|>system
List of tools: <|tool_list_start|>[{"name": "get_candidate_status", "description": "Retrieves the current status of a candidate in the recruitment process", "parameters": {"type": "object", "properties": {"candidate_id": {"type": "string", "description": "Unique identifier for the candidate"}}, "required": ["candidate_id"]}}]<|tool_list_end|><|im_end|>
<|im_start|>user
What is the current status of candidate ID 12345?<|im_end|>
<|im_start|>assistant
<|tool_call_start|>[get_candidate_status(candidate_id="12345")]<|tool_call_end|>Checking the current status of candidate ID 12345.<|im_end|>
<|im_start|>tool
<|tool_response_start|>[{"candidate_id": "12345", "status": "Interview Scheduled", "position": "Clinical Research Associate", "date": "2023-11-20"}]<|tool_response_end|><|im_end|>
<|im_start|>assistant
The candidate with ID 12345 is currently in the "Interview Scheduled" stage for the position of Clinical Research Associate, with an interview date set for 2023-11-20.<|im_end|>

You can directly pass tools as JSON schema or Python functions with .apply_chat_template() as shown in this page to automatically format the system prompt.

Architecture: Hybrid model with multiplicative gates and short convolutions: 10 double-gated short-range LIV convolution blocks and 6 grouped query attention (GQA) blocks.

Pre-training mixture: Approximately 75% English, 20% multilingual, and 5% code data sourced from the web and licensed materials.

Training approach:

• Knowledge distillation using LFM1-7B as teacher model
• Very large-scale SFT on 50% downstream tasks, 50% general domains
• Custom DPO with length normalization and semi-online datasets
• Iterative model merging

🏃 How to run LFM2

1. Transformers

To run LFM2, you need to install Hugging Face transformers v4.55 or a more recent version as follows:

pip install -U transformers

Here is an example of how to generate an answer with transformers in Python:

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model_id = "LiquidAI/LFM2-350M"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype="bfloat16",
#    attn_implementation="flash_attention_2" <- uncomment on compatible GPU
)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Generate answer
prompt = "What is C. elegans?"
input_ids = tokenizer.apply_chat_template(
    [{"role": "user", "content": prompt}],
    add_generation_prompt=True,
    return_tensors="pt",
    tokenize=True,
).to(model.device)

output = model.generate(
    input_ids,
    do_sample=True,
    temperature=0.3,
    min_p=0.15,
    repetition_penalty=1.05,
    max_new_tokens=512,
)

print(tokenizer.decode(output[0], skip_special_tokens=False))

# <|startoftext|><|im_start|>user
# What is C. elegans?<|im_end|>
# <|im_start|>assistant
# C. elegans, also known as Caenorhabditis elegans, is a small, free-living
# nematode worm (roundworm) that belongs to the phylum Nematoda.

You can directly run and test the model with this Colab notebook.

2. vLLM

You need to install vLLM v0.10.2 or a more recent version as follows:

uv pip install vllm==0.10.2 --extra-index-url https://wheels.vllm.ai/0.10.2/ --torch-backend=auto

Here is an example of how to use it for inference:

from vllm import LLM, SamplingParams

prompts = [
    "What is C. elegans?",
    "Say hi in JSON format",
    "Define AI in Spanish"
]
sampling_params = SamplingParams(
    temperature=0.3,
    min_p=0.15,
    repetition_penalty=1.05
)

llm = LLM(model="LiquidAI/LFM2-350M")

outputs = llm.generate(prompts, sampling_params)

for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")

3. llama.cpp

You can run LFM2 with llama.cpp using its GGUF checkpoint. Find more information in the model card.

🔧 How to fine-tune LFM2

We recommend fine-tuning LFM2 models on your use cases to maximize performance.

Notebook	Description	Link
SFT (Unsloth)	Supervised Fine-Tuning (SFT) notebook with a LoRA adapter using Unsloth.
SFT (TRL)	Supervised Fine-Tuning (SFT) notebook with a LoRA adapter using TRL.
DPO (TRL)	Preference alignment with Direct Preference Optimization (DPO) using TRL.

📈 Performance

LFM2 outperforms similar-sized models across different evaluation categories.

1. Automated benchmarks

Model	MMLU	GPQA	IFEval	IFBench	GSM8K	MGSM	MMMLU
LFM2-350M	43.43	27.46	65.12	16.41	30.1	29.52	37.99
LFM2-700M	49.9	28.48	72.23	20.56	46.4	45.36	43.28
LFM2-1.2B	55.23	31.47	74.89	20.7	58.3	55.04	46.73
Qwen3-0.6B	44.93	22.14	64.24	19.75	36.47	41.28	30.84
Qwen3-1.7B	59.11	27.72	73.98	21.27	51.4	66.56	46.51
Llama-3.2-1B-Instruct	46.6	28.84	52.39	16.86	35.71	29.12	38.15
gemma-3-1b-it	40.08	21.07	62.9	17.72	59.59	43.6	34.43

2. LLM-as-a-Judge

3. Inference

Throughput comparison on CPU in ExecuTorch

Throughput comparison on CPU in Llama.cpp

📬 Contact

If you are interested in custom solutions with edge deployment, please contact our sales team.

Citation

@article{liquidai2025lfm2,
 title={LFM2 Technical Report},
 author={Liquid AI},
 journal={arXiv preprint arXiv:2511.23404},
 year={2025}
}